One AI assistant for DingTalk, Feishu and WeCom

Many companies use two or three collaboration platforms at once: headquarters on Feishu, a subsidiary or business unit on DingTalk, and WeCom for talking to customers. When they bring in an AI assistant, the first instinct is to set one up on each platform. Then three assistants each keep their own knowledge, permissions and workflows, maintenance multiplies, and the same question can get different answers on different platforms.

A simpler approach: run one AI assistant on a single platform. Connect the three collaboration tools as entry points and notification channels, and connect business systems behind it as tools.

1. Collaboration platforms are entry points, not the assistant

An employee sends a request in DingTalk, Feishu or WeCom. The message goes to the same assistant, and the result returns to the platform it came from. The assistant's knowledge base, the systems it can call and the approval rules are maintained once, and every platform follows the same rules.

Switching or adding a collaboration platform then means adding one more entry point, not rebuilding the assistant.

2. Business systems connect behind it as tools

An AI assistant is useful because it can look things up and get things done: check stock in the ERP, look up a customer in the CRM, submit an approval in the OA system. These systems connect to the platform as tools, so whichever collaboration platform a request comes from, it calls the same set of tools.

Before connecting, list which systems you need and whether each must be read or also written. For the main types of platform and how they connect systems, see choosing an enterprise agent platform: self-hosted, SaaS and low-code.

3. Identities and permissions must line up

One employee has three accounts across DingTalk, Feishu and WeCom, but to the assistant they should be one person with one set of permissions. Otherwise something visible on one platform is hidden on another, or the assistant ends up with more access than the person has. Before going live, map the three platforms' accounts to the company directory and set permissions per person, not per platform.

4. Write operations wait for confirmation

When the assistant needs to change data, submit an approval or send a message, it first hands the action to a person to confirm, runs it only after confirmation, and keeps a record. Write operations from all three platforms go through the same confirmation step, so there is one consistent rule.

5. Start with one platform and one task

There is no need to connect all three platforms at once. Pick the platform most people use and one routine task that spans two systems with fixed steps, for example receiving a request in the collaboration tool, looking up a business system, and submitting the result for approval. Once that works, add the other two platforms as entry points.

Summary

With DingTalk, Feishu and WeCom all in use, you need one AI assistant: collaboration platforms as entry points, business systems as tools, identities matched per person, and write operations confirmed by a person first. Get one task working on one platform, then extend to the rest.

For reference, FIM One from Beijing FIM Network Technology Co., Ltd. (北京飞沐网络科技有限公司) is a self-hosted enterprise agent platform for companies that use global SaaS alongside DingTalk, WeCom and Feishu. It works in three steps: connect existing systems, orchestrate across them, and have a person confirm write operations before they run. FIM One is source-available software, and the Community edition is free forever and self-hosted. See the connectors for how systems are connected, and where human review sits for how write operations are confirmed.